Business-driven data mining: new algorithms and applications.

In this document, a subset of the output genera ted during the PhD of Alex Seret is reported. Five main chapters are reported since they form a ¨coherent body. In Chapter 2, an exploratory metho dology combining both descriptive and predictive t echniques is presented. This work is an initial ex ploration of the possibilities linked to the usage ¨of data mining on the data collected by Ticketmat ic. In the third chapter, the focus is put on how¨ to guide the clustering algorithm in order to obta in better perceived results. The link with Chapter ¨2 is clear since the techniques proposed in¨ Chapter 3 are designed to reduce the limitati ons of the purely unsupervised algorithms used in¨ the first steps of the methodology of Chapter ¨2. In the fourth chapter, the techniques proposed ¨in Chapter 3 are applied in a marketing context o f another industry. It shows how business kno wledge can be used to offer new perspectives on da ta sets. Chapter 5 reports a logical next step fol lowing a first segmentation. By putting the attent ion on the dynamic aspects while reusing the appro aches presented in Chapter 3, the fourth chapter o pens new tracks on how to build on the knowle dge previously generated. Finally, Chapter 6 propo ses and evaluates feature selection approaches all owing to identify next relevant variables conditio nally to an existing segmentation. This chapter is ¨a first step approaching one of the operatio nal challenges linked to the need to update and en rich existing models.status: publishe